Stock prediction using different machine learning techniques with the dataset of NVIDIA
Date
2025
Authors
Chen, Ying
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This project investigates the effectiveness of machine learning models in predicting NVIDIA Corporation's stock price movements. Four distinct ML approaches were implemented and compared: LR (Linear Regression), SVM (Support Vector Machines), NN (Neural Networks), and LSTM (Long Short-Term Memory networks). Using historical price data and technical indicators, model performance was evaluated through RMSE (Root Mean Squared Error) and R² (R-squared) metrics.
The results demonstrated that LR and SVM models performed better than complex deep learning architectures, achieving superior accuracy, with RMSE ≈ 3, R² ≈ 0.99, while maintaining interpretability. The Neural Network model performed unexpectedly and poorly, with R² = -0.50, suggesting significant overfitting challenges. While the LSTM showed promise for capturing temporal dependencies, it required further optimization to achieve higher accuracy and compete with traditional methods.
In addition, the project incorporated the importance of ethical considerations, including bias mitigation strategies and regulatory compliance measures for financial AI applications. The critical balance between model complexity and practical utility was highlighted in stock market prediction, emphasizing that sophisticated architectures don't automatically guarantee better performance.
Description
Keywords
stock price prediction, machine learning, algorithmic trading, financial forecasting, AI ethics